LKE/KDSL Research Seminar

Christian Wirth and Oliver Ferschke will present their work at the LKE/KDSL Research Seminar.

On July 2, 2013, at 11:40am in S1|03 223, we are honored to hold the monthly LKE Research Seminar with two speakers, as follows:

Christian Wirth will present:

Preference-based Reinforcement Learning

Reinforcement learning is a generic machine learning approach for solving problems which can be modeled as sequence of states and actions. This approach is rarely applied in NLP domains, because it is not trivial to model those problems as a reinforcement learning problem. Aspects like the rewards or the Markov property have to be considered. Preference-based approaches are trying to get rid of some of those limitations by introducing pairwise preference relations as a replacement for the reward function. This is especially useful in domains where numeric rewards are dependent on a human evaluation, resulting in unreliable values. This talk will give a short overview of reinforcement learning in general, as well as some examples of its application in NLP, followed by an introduction to preference based approaches.

Oliver Ferschke will present:

The Impact of Topic Bias on Quality Flaw Prediction in Wikipedia

With the increasing amount of user generated reference texts in the web, automatic quality assessment has become a key challenge. However, only a small amount of annotated data is available for training quality assessment systems. Wikipedia contains a large amount of texts annotated with cleanup templates which identify quality flaws. We show that the distribution of these labels is topically biased, since they cannot be applied freely to any arbitrary article. We argue that it is necessary to consider the topical restrictions of each label in order to avoid a sampling bias that results in a skewed classifier and overly optimistic evaluation results. We factor out the topic bias by extracting reliable training instances from the revision history which have a topic distribution similar to the labeled articles. This approach better reflects the situation a classifier would face in a real-life application.